Mattia Gaggi

Developed 3D lung nodule segmentation models on CT imaging data at Optellum, a spin-off from the Oxford University Visual Geometry Group (VGG). Contributed to one of the first FDA-approved AI systems for early lung cancer diagnosis, deployed across hospitals in the United States.

• Improved classification AUC by ~1% — in medical diagnostics even small AUC improvements significantly improve detection accuracy
• Developed 3D segmentation models for identifying lung nodules in CT scan data
• Models deployed in production across US hospital systems

Designed and improved ML models for automatically classifying and interpreting insurance documents at Tractable, one of Europe's most prominent AI companies. Built transformer-based NLP systems (RoBERTa) and iteratively improved the full ML pipeline including document segmentation, OCR integration, and named entity recognition.

• Retained approximately 45% of company's Annual Recurring Revenue
• Contributed to internal cost savings estimated at over $3M per year
• Expanded Tractable's services into North America through robust document processing
• Improved system robustness across heterogeneous documents from different countries and formats

Led development of machine learning recommendation systems for in-game items used by millions of players at a global mobile gaming company. Designed and implemented a full production pipeline using Airflow for generating personalised recommendations, targeting mid-level players most at risk of churning.

• Achieved 1% improvement in engagement for the target player segment — significant at millions-of-users scale
• Designed end-to-end ML pipelines from training through production serving
• Played a strategic role shaping the team's ML direction as one of two ML engineers

Founding engineer at Morph Systems, an agentic AI platform that enables users to create complex software applications using coordinated AI agents. Designed core system architecture working directly with senior technical leadership including an ex-Google Director of Engineering.

• Designed and implemented multi-agent orchestration architecture for collaborative software development tasks
• Built distributed pipelines for coordinating agents that interact with APIs, development tools, and external systems
• Implemented scalable infrastructure for AI agent task execution across cloud environments

Built a generative AI platform that converts text descriptions into printable 3D models during the Entrepreneur First accelerator programme — a leading global deep-technology accelerator that has produced multiple billion-dollar companies.

• Reached 1,000 monthly active users shortly after launch
• Built generative pipeline using Stable Diffusion and NeRF models to synthesise 3D geometry from text
• Designed distributed GPU cluster infrastructure and Discord-based user interface

Moonsift — Tech Lead (contract). Defined technical roadmap for AI recommendation system. Built multimodal search models and hybrid ranking systems. Improved system speed by 50%.
AutoEnhance.ai — Head of AI. Developed computer vision models for automated real-estate photo editing including multi-exposure fusion, color correction, and segmentation.
Intelligent Datalytics — Data Scientist Consultant. Built ML solutions for pharmaceutical and telecommunications clients, including a churn prediction model for 18M customers achieving 3% churn reduction.